NLMDB: Natural Language & MCP-powered Database

Query your databases using natural language through the Model Context Protocol (MCP) approach. NLMDB provides a simple API for interacting with databases using either OpenAI or Hugging Face models.

✨ Features

• 💬 Query databases using natural language
• 🔄 Support for both OpenAI and Hugging Face models
• 🔒 Enhanced privacy options with local Hugging Face models
• 📊 Automatic schema extraction
• 📈 Generate visualizations from your data
• 📝 Multiple output formats: explanatory text, raw data, or visualizations
• 🧩 Simple, intuitive API

🚀 Installation

pip install nlmdb

🏁 Quick Start

Natural Language Explanations Mode

from nlmdb import dbagent

# Initialize the agent with your API key and database path
response = dbagent(
    api_key="your-openai-api-key",
    db_path="path/to/your/database.db",
    query="What tables are in the database and what columns do they have?"
)

print(response["output"])

SQL Agent Mode

Get direct results without explanations - perfect for data analysis workflows:

from nlmdb import sql_agent
import pandas as pd

# Get results as a pandas DataFrame
df = sql_agent(
    api_key="your-openai-api-key",
    db_path="path/to/your/database.db",
    query="List all customers who made purchases over $1000",
    return_type="dataframe"  # Options: "dataframe", "dict", or "json"
)

# Now you can directly work with the data
print(df.head())

Visualization Agent Mode (New!)

Generate interactive visualizations directly from your database with natural language:

from nlmdb import viz_agent

# Create a visualization
fig = viz_agent(
    api_key="your-openai-api-key",
    db_path="path/to/your/database.db",
    query="Show me a bar chart of sales by product category"
)

# Display the interactive plot
fig.show()

# Save the plot to HTML
fig.write_html("sales_by_category.html")

Using Hugging Face Models

from nlmdb import dbagent_private

# Initialize the agent with your Hugging Face token and model name
response = dbagent_private(
    hf_config=("your-huggingface-token", "model-repo-name"),
    db_path="path/to/your/database.db",
    query="What tables are in the database and what columns do they have?"
)

print(response["output"])

🔒 Privacy and Data Security

NLMDB offers enhanced privacy options through its support for Hugging Face models:

Enhanced Privacy with Hugging Face Models

When using dbagent_private, sql_agent_private, or viz_agent_private with use_local=True, all processing happens locally on your machine, ensuring your database schema and query data never leave your environment:

response = dbagent_private(
    hf_config=("your-huggingface-token", "model-repo-name"),
    db_path="path/to/your/database.db",
    query="What tables are in the database?",
    use_local=True  # Ensures all processing happens locally
)

Data Security Considerations

• OpenAI Integration: When using dbagent with OpenAI models, database schema and queries are sent to OpenAI's API. While only schema information and not actual data is shared, consider privacy implications.

• Hugging Face Cloud API: Using dbagent_private without use_local=True sends queries to Hugging Face's Inference API.

• Local Processing: For maximum privacy, use dbagent_private with use_local=True to keep all processing on your machine.

• No Data Storage: NLMDB does not store or log your database contents, queries, or responses.

🔄 Model Comparison

🧩 Advanced Usage

Visualization Agent with Different Output Formats

# Get HTML output for embedding in web applications
html_output = viz_agent(
    api_key="your-openai-api-key",
    db_path="path/to/your/database.db",
    query="Create a pie chart showing the distribution of sales by region",
    return_fig=False,
    fig_format="html"
)

# Save to an HTML file
with open("sales_distribution.html", "w") as f:
    f.write(html_output)

# Get JSON specification for further customization
json_spec = viz_agent(
    api_key="your-openai-api-key",
    db_path="path/to/your/database.db",
    query="Show me a scatter plot of product price vs. sales volume",
    return_fig=False,
    fig_format="json"
)

SQL Agent with Different Return Types

# Get results as a dictionary
result_dict = sql_agent(
    api_key="your-openai-api-key",
    db_path="path/to/your/database.db",
    query="Find the total sales by product category",
    return_type="dict"
)

# Get results as JSON
json_result = sql_agent(
    api_key="your-openai-api-key",
    db_path="path/to/your/database.db",
    query="Show me monthly sales trends",
    return_type="json"
)

# SQL Agent with Hugging Face for privacy
df = sql_agent_private(
    hf_config=("your-huggingface-token", "mistralai/Mixtral-8x7B-Instruct-v0.1"),
    db_path="path/to/your/database.db",
    query="List customers in California",
    return_type="dataframe",
    use_local=True  # For local processing
)

Customizing Model Parameters

model_kwargs = {
    "temperature": 0.2,
    "max_new_tokens": 1024,
    "repetition_penalty": 1.1
}

response = dbagent_private(
    hf_config=("your-huggingface-token", "mistralai/Mixtral-8x7B-Instruct-v0.1"),
    db_path="path/to/your/database.db",
    query="Summarize the sales data for the last quarter",
    model_kwargs=model_kwargs
)

🔍 Choosing the Right Mode & Model

Modes

Recommended Hugging Face Models

📊 Supported Databases

Currently, NLMDB supports:

• SQLite ✅

Future releases will add support for:

• PostgreSQL 🔜
• MySQL 🔜
• Microsoft SQL Server 🔜

⚙️ Requirements

• Python 3.8+
• openai>=1.0.0
• langchain>=0.1.0
• langchain-core>=0.1.0
• langchain-community>=0.0.0
• langchain-huggingface>=0.0.1 (for Hugging Face integration)
• pandas>=1.0.0 (for DataFrame return type in SQL agent mode)
• plotly>=5.0.0 (for visualization agent mode)

📜 License

MIT

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add some amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

🙏 Acknowledgements

This library is built on top of:

• LangChain
• OpenAI API
• Hugging Face Inference API
• SQLite
• pandas
• Plotly